A New Genetic Algorithm for Overlapping Community Detection

Abstract

In this paper, we proposed a new genetic algorithm for detecting overlapping communities in various, real-world social networks. This algorithm uses a new strategy to transform the adjacency matrix by weighting edges according to their centrality. Also, we modified some rules so that genetic algorithms can be used to detect overlapping communities. Matrix-based encoding was used to improve the efficiency of crossover and decoding operations. We used the similarity of nodes to generate the initial populations. However, when a node belonged to different communities, we decided whether it should be assigned to these communities by using threshold determination. Thus, we detect overlapping communities. By experimental comparison with other algorithms, the simulation results showed that the WBDF algorithm was effective and efficient in the detection of overlapping communities.